A systematic study of the convergence of QM/MM results with respect to the chosen size of the QM region is presented for two examples of peptidic systems. For this purpose, we increased the QM region to up to 1637 atoms at the HF/SVP and 383 atoms at the SOS-AO-MP2/6-31G** level. While the convergence behavior is almost independent of the chosen method and basis set, the study clearly shows that for the considered proton-transfer energy the QM/MM treatment leads to a significantly faster convergence than the pure QM treatment. This behavior can be rationalized by the fair description of the surrounding of the active center using MM methods, even though the MM description of the active center is not adequate in our present case. At the same time, the observed convergence is-quite insensitive to a variation of charge surroundings in the chosen model peptides. Although the QM/MM results do converge much quicker with the system size than the pure QM ones, the data show that even for the chosen simple model systems about 150-300 QM atoms are needed to achieve accuracies in the order of 10 kJ/mol and about 300-1000 atoms for an accuracy of 2 kJ/mol with respect to a convergence with the QM-region size.